Lipids, as proteins and nucleic acids, make up the building blocks of living cells, serve as important energy sources or intra- or extracellular signaling molecules [1, 2]. During cancer development and progression, cancer cells need to constantly adapt metabolically to tumor microenvironment (TME) evolutions, such as limited vascularization and nutrient bioavailability, as well as increased extracellular matrix (ECM), immune infiltrate and cancer-associated fibroblasts (CAFs) expansion, in order to grow, proliferate and form metastases [3, 4]. Hence, depending on tissue of origin, oncogenic drivers and exogenous nutrient availability, cancer cells may rely on lipid uptake or on intracellular lipid pools, resulting from de novo synthesis, lipid droplets (LD) mobilization (e.g., lipolysis, cholesterol de-esterification), or membrane remodeling [5,6,7,8]. Moreover, in TME-rich tumors, recruitment of non-malignant host cells, such as immunosuppressive cells, CAFs, cancer-associated adipocytes (CAAs), can be induced by tumor- or stromal-derived lipid mediators (e.g., Prostaglandin E2 (PGE2), fatty acids (FAs), cholesterol) and the acquisition of pro-tumoral functions of TME’s cells depends in part of lipid reprogramming [9,10,11]. Therefore, having an overview of lipid functions in malignant and non-malignant compartments, and in the dialog between both is absolutely needed before considering systemic treatment targeting lipid metabolism as therapeutic strategies for cancer. Moreover, there are growing evidences from preclinical cancer studies that activated lipid metabolic pathways in tumoral and/or TME compartments participate to the increased cancer cell resistance to chemotherapeutic agents or immunotherapy [9, 12]. Therefore, the targeting of lipid metabolism could also be a promising therapeutic approach to overcome tumor resistance to most common treatments.

In this review, we summarize the main findings showing how lipid and cholesterol metabolism in the tumor and stromal compartments supports metastases formation. As their respective role in parental tumor growth has been stated in several recent reviews [4, 6,7,8, 13], this aspect will not be addressed in the review. Here, we discuss the role of tumor-derived lipids, as signaling molecules promoting pro-tumoral TME, and how the contribution of TME-specific lipid reprogramming is beneficial for the tumors. Finally, we summarize therapeutic strategies in which targeting lipid pathways can improve tumor response to chemo- or immunotherapy.

Targeting lipid metabolism to slow down metastatic disease

Metastases for all solid tumors account for 66.7% of cancer deaths [14], and their formation is a multistep process starting at the primary site, then pursuing in the vasculature to finally ending in distant organs [15]. Primary cancer cells need to increase their motility and invasive capacities, through the ECM, to reach the local vasculature and then the circulation [15]. This is accompanied by an epithelial to mesenchymal transition (EMT), during which epithelial cells acquire a mesenchymal phenotype which is prone to colonization of distant sites. However, our understanding of the metabolic changes characterizing each step ultimately leading to the colonization of distant healthy tissues is still far behind that of parental primary tumor [16]. Moreover, according to the metastatic site, cancer cells have also to cope with changes in nutrient and oxygen intakes to survive and expand [16]. Hence, changes in their metabolic program, as well as increased metabolite exchanges with the surrounding non-malignant cells are key determinants of metastatic colonization and growth [17].

Role of fatty acid (FA) transport in the metastatic process

FAs are either provided by diet or directly synthetized through the de novo synthesis pathway (Fig. 1). They are used as energetic sources to fuel tricarboxylic acid (TCA) cycle or as substrates for the synthesis of complex lipids, including structural lipids (e.g., phospholipids (PLs), sphingolipids) and intra- or extracellular signaling molecules (e.g., lysophospholipids, prostaglandins, leukotrienes) [1, 2] (Fig. 1). Different membrane-associated proteins are involved in the uptake of exogenous FAs by cancer cells from their environment, such as FA translocase (FAT or CD36), and FA transport protein (FATP) or in the intracellular transport of FAs, such as plasma membrane FA-binding protein (FABP) [7]. Nieman et al. have demonstrated that, during early stages of metastatic cascade, FABP4 promotes detachment and migration of cancer cells through the ECM towards the vasculature and is the core of the dialog between metastatic ovarian cancer (OvCa) cells and CAAs [18]. In this context, intracellular long-chain FA transport by FABP4, which is upregulated in patient’s omental metastases as compared to primary tumor, increases invasive capacities of OvCA cells [18, 19]. However, even if FABP4 is detected both in OvCA cells and CAAs, systemic treatments targeting tumor and host FABP4 are not more effective in decreasing metastatic nodules than host FABP4 knock-down [19, 20], highlighting the prominent role of FABP4 in TME [18,19,20]. In clinic, FABP4high patients show poor overall and progression-free survival, and can be discriminated from FABP4low patients by the lipid profile of OvCa cells, which is enriched in glycerolipids, PL and LysoPLs and in unsaturated and oxidized-FA lipid species [19]. In addition to FABP4, CD36 appears as an interesting metabolic target to limit metastatic progression in cancer. Indeed, its inhibition in OvCA cells restrains accumulation of LDs, cholesterol and lipid peroxidation products and diminishes their invasive and migratory capacities, as well as their adhesion on most ECM components of the peritoneum [21]. Consequently, intraperitoneal injection of CD36 KD cells gives rise to few metastases, and in an encouraging way, similar results are obtained after daily injections of neutralizing anti-CD36 antibody in wild-type (WT) OvCA xenograft models [21]. Inversely, CD36 overexpression in human oral carcinoma cells with low metastatic potential drives macro-metastasis in lymph nodes without potentiating primary tumor size [22]. This metastatic phenotype depends on increased expression of fatty acid oxidation (FAO)-related genes, and is boosted by a high-fat diet providing an enhanced supply of FAs to cancer cells. Interestingly, CD36-neutralizing antibody is equally effective on metastatic regression in intravenously-inoculated melanoma, oral carcinoma and mammary gland cancer mouse models [22]. Finally, a positive correlation between CD36 and FABP expression and EMT markers is observed in hepatocellular carcinoma (HCC) patients suffering of obesity [23]. Likewise, supplementation with palmitic or oleic acid improves HCC cell migration and activates the TGF-β and Wnt signaling pathways, both responsible of EMT program [23]. However, whether these effects are truly dependent on CD36-mediated FA uptake, still remains to be determined. Together, these studies provide evidences that FA transport is intimately connected to metastatic progression, at least in some cancers.

Fig. 1: Non-exhaustive representation of lipid metabolic pathways addressed in the review.

Fatty acids (FAs), taken up from tumor microenvironment through FA translocase (CD36), FA transport protein (FATP) or by other mechanisms (e.g., passive diffusion, lipoprotein endocytosis), are esterified by acyl-CoA synthetases (ACS) into fatty acyl-CoA (FA-CoA). These latter are used either to build structural or storage lipids or as energy sources through the fatty acid β-oxidation (FAO) pathway. a Fatty acid β-oxidation. The first and rate-limiting step of long-chain acyl-CoAs (LCFA-CoAs) oxidation is accomplished by the carnitine palmitoyltransferase 1 (CPT1) located in the outer mitochondrial membrane. CPT1 converts LCFA-CoA into LCFA-carnitine, a necessary step to provide the LCFA-CoA entry into the mitochondrial matrix, whereas short- and medium-chain FA-CoAs (S- and MCFA-CoAs)) can freely diffuse through the mitochondrial inner membrane. Then, a carnitine/acylcarnitine translocase shuttles the LCFA-carnitine across the inner mitochondrial membrane where they are converted back to LCFA-CoAs by CPT2. Unlike saturated FA-CoA, polyunsaturated FA-CoAs (PUFA-CoAs) need to be sequentially converted in trans-2-enoyl-CoA by the 2,4-dienoyl-CoA reductase (DECR) and the Δ3, Δ2-enoyl-CoA isomerase (ECI) enzymes, before to be completely degraded through the β-oxidation cycle. The latter, through the successive action of the acyl-CoA dehydrogenase (ACAD) and the α and β-subunits of trifunctional protein TP, gives rise to acetyl-CoA molecules, which are further oxidized through the tricarboxylic acid (TCA) cycle to generate ATP and the reducing equivalents NADH and FADH2. b De novo lipogenesis (DNL). Citrate, derived from the TCA cycle, is converted to acetyl-CoA into the cytoplasm by the ATP-citrate lyase (ACLY). Then, acetyl-CoA is carboxylated by the cytoplasmic acetyl-CoA carboxylase 1 (ACC1) into malonyl-CoA, which is considered as the rate-limiting reaction. The next step is catalyzed by the fatty acid synthase (FAS), that synthesizes one molecule of palmitic acid from 7 molecules of malonyl-CoA. Further elongation and insertion of double bonds in carbon chains from FA (provided by DNL and/or diet), are carried out by FA elongases (ELOVLs) and stearoyl-CoA or FA desaturases (SCDs or FADSs), respectively. Glutamine and acetate are also suppliers of citrate and acetyl-CoA useful for DNL. c Eicosanoid synthesis. The arachidonic acid (ARA), released from membrane phospholipid (PL) hydrolysis by the cytoplasmic phospholipase A2α (cPLA2α), is further metabolized into either prostaglandins (PG) and thromboxanes (TX), or leukotrienes (LT) and hydroxyeicosatetranoic acids (HETE). Cyclooxygenases (COX) and lipoxygenases (LOX) constitute the respective rate-limiting enzymes of PG and LT/HETE synthesis, respectively. All eicosanoids are secreted and thereby exert autocrine and/or paracrine functions. d Neutral lipolysis. Hydrolysis of triacylglycerols (TAG) to FA and diacyglycerol (DAG) is driven by the adipose triglyceride lipase (ATGL), while the DAG breakdown, producing monoacylglycerol (MAG) and FA, is catalyzed by the hormone-sensitive lipase (HSL). Finally, the monoacylglycerol lipase (MAGL) hydrolyses MAG into glycerol and FA. SSO sulfo-N-succinimidyl oleate, ETC electron transport chain, MUFA monounsaturated FA, lysoPL lysophospholipid, 5-HPETE 5-hydroperoxyeicosatetraenoic acid. LPCAT Lysophosphatidylcholine acyltransferase, GPAT glycerol-3-phosphate acyltransferase, ACSS2 ACS short-chain family member 2, TOFA 5-Tetradecyloxy-2-furoic acid, NDGA nordihydroguaiaretic acid, PTGES Prostaglandin E synthase, 15-PGDH 15-hydroxyprostaglandin dehydrogenase. Indirect and direct chemical reactions are illustrated by dotted and solid arrows, respectively.

Catabolic pathways promoting metastasis Fatty acid oxidation (FAO)

The initial substrates for FAO are acyl-CoAs, which result from the esterification of FAs by acyl-CoA synthetases (ACS) (Fig. 1). The resulting acyl-CoAs, which cannot be exported from cells, are then broken down into acetyl-CoA molecules, whose number depends upon the carbon chain length of the acyl-CoA being oxidized. Finally, FAO-derived acetyl-CoA is further oxidized through the TCA cycle, which in cooperation to oxidative phosphorylation, ultimately generates ATP and reducing equivalents, NADH and FADH2, to support redox homeostasis.

It is increasingly evident that, besides its undisputed role in tumor growth promotion [6, 7, 13], FAO is essential to support survival of ECM-detached cancer cells and their establishment in the metastatic niche [24, 25]. Indeed, in condition of loss of attachment, melanoma cells over-activate FAO to maintain high levels of NADPH and thereby prevent ROS-induced cell death [24]. Mechanistically, the FAO rate-limiting enzyme (i.e., mitochondrial trifunctional protein (TPβ)) (Fig. 1) is constitutively activated by the binding of phosphorylated Nur77 in the mitochondria. Hence, knockdown of TPβ or Nur77 reduces the number of circulating tumor cells and consequently the risk of lung colonization by these cells [24]. Interestingly, this is validated in patients with metastatic melanoma, who show higher Nur77 levels in metastases than in primary tumors [24]. Therefore, disrupting Nur77-TPβ interaction may constitute a therapeutic strategy for these metastatic patients. Another study demonstrated that melanoma cells undergo a metabolic shift from glucose/glutamine oxidation in primary tumors towards FAO in lymph node metastases to meet their bioenergetic needs [25]. Indeed, the higher increase in FAO gene transcription in lymph node-metastasis as compared to primary tumor and lung metastases results from activation of YAP signaling mediated by the binding of bile acids on vitamin D receptors [25]. Hence, administration of bile acids, generated from cholesterol in a CYP7A1 manner, promotes the growth of lymph node-metastatic melanoma, while depletion of CYP7A1 has an opposite effect [25]. Therefore, targeting YAP and/or CYP7A1 may be interesting therapeutic options to prevent the risk of cancer cell dissemination towards lymph nodes.

Neutral lipolysis

Triacylglycerols (TAG) constitute the storage form of excess intracellular FAs in LDs [26]. Their hydrolytic breakdown by 3 consecutive cytosolic lipases, i.e., neutral lipolysis, gives rise to FAs and glycerol (Fig. 1) [27]. Several studies showed the impact of blocking this metabolic route on metastasis formation, either directly by acting on lipolysis-mediated enzymes (e.g., monoacylglycerol lipase (MAGL) and hormone-sensitive lipase (HSL)), or indirectly on long-chain ACS (ACSL), promoting long-chain FA activation [28,29,30] (Fig. 1). Indeed, enhanced MAGL activity is a common feature of several aggressive cancers (i.e., melanoma, breast cancer (BC), OvCA) [28]. Its genetic or pharmacological inhibition leads to an accumulation of monoacylglycerols and subsequent reduction in FAs. These metabolic changes are accompanied by reduced tumorigenic capacities both in vitro and in vivo, which are fully rescued by supplementation in saturated fatty acids or high-fat diet, respectively [28]. Also in aggressive triple-negative breast cancer (TNBC), blocking the increased lipolysis induced by the cleaved-form of CUB-domain containing protein 1 (CDCP1), a feature also found in several other cancers (e.g., ovarian, lung, colon, prostate, and pancreatic cancers), decreases the metastatic abilities of TNBC cells in vitro and in vivo [30]. Mechanistically, disruption of CDCP1-ACSL interaction promotes activation of long-chain-FAs, and their subsequent degradation through the FAO and TCA cycle [30]. On the contrary, in Kras-driven pancreatic ductal adenocarcinoma (PDAC), HSL is down-regulated as compared to normal pancreas, and the subsequent decrease in HSL-induced lipolysis promotes the metastatic potential of mutated-KRAS PDAC cells [29]. This pro-metastatic phenotype is rescued by oleic acid supplementation, which restores intracellular LD pool. Interestingly, oleic acid primes PDAC cells to ECM degradation, invadopodia formation, as well as to migration and invasion. Consequently, HSL-expressing KrasG12D PDAC cells are unable to form metastases neither primary tumor as compared to control KrasG12D PDAC cells [29]. Importantly, these data highlight the therapeutic benefits of direct and indirect targeting of lipolysis on metastatic progression, especially for aggressive cancers.

Activated anabolic pathways supporting metastasis De novo lipogenesis (DNL)

DNL, also known as de novo FA synthesis, is the metabolic pathway by which FAs are generated from different carbon sources (Fig. 1). In some cancers, this anabolic process is activated [4, 7, 31], and acetyl-CoA, produced from glucose catabolism in well-oxygenated condition or from alternative carbon sources, glutamine or acetate, under metabolic stress [32, 33], provides the substrate to DNL (i.e., citrate) (Fig. 1). Interestingly, in BC, non-small cell lung cancer (NSCLC) and prostate cancer (PCa), the relative contribution of de novo synthetized FAs from glucose and glutamine in the intracellular lipid pool is much lower than that of extracellular FAs (5–35% vs 65–75%) [7]. However, even if a therapeutic advantage of interrupting DNL on tumor growth has been demonstrated in several preclinical cancer models, the clinical development of more efficient, selective and safe DNL inhibitors is needed [34, 35]. Importantly, two recent studies showed that basal-like and HER2+ BC cells have a higher DNL rate when they grow in brain metastatic site rather than in primary one, and the DNL signature, characterized by increased cholesterol species and structural lipids, governs their ability to metastasize exclusively in the brain [36, 37]. These results are in accordance with clinic findings showing an up-regulation of SREBF1 and FASN in brain metastasis relative to extra-cranial metastasis or matched-primary tumors in metastatic BC patients [36, 37]. Interestingly, SREBF1-depleted BC cells still able to seed but not to expand in brain microenvironment [36]. Similarly, the brain-permeable FAS inhibitor (i.e., BI99179) (Fig. 1) limits the growth of BC metastases in the brain, but does not slow primary tumor growth [37]. In the context of Pten null PCa where DNL contributes to the increase of TAG and PL pools, targeting of the SREBP-mediated lipogenic program using fatostatin also impedes distant lymph node metastasis, as well as primary tumor growth by decreasing the frequency of mitotic cancer cells and activating apoptosis [38]. Thus, these findings demonstrate that enhanced DNL pathway is a common feature of BC and PCa metastasis that can be targeted to slow down disease progression and expansion.

Prostaglandin and leukotriene synthesis

The released arachidonic acid (ARA) from membrane PLs, by the cytosolic phospholipase A2 alpha (cPLA2α), is further metabolized by 2 rate-limiting enzymes into active metabolites, namely eicosanoids (Fig. 1). The cyclooxygenases (COX) generate prostaglandins (PG), prostacyclins and thromboxanes, while the lipoxygenases (LOX) are responsible of leukotriene (LT) and hydroxyeicosatetraenoic acid (HETE) production [39] (Fig. 1). Once secreted, they mediate tumor promotion and progression by acting on producer or neighboring cells in an autocrine or paracrine manner, respectively [39] (Fig. 3). Regarding lung cancer, PGD2 to PGF2 are synthetized by both lung cancer cells and TME cells, while LTB4 to LTE4 are uniquely produced by tumor-associated-neutrophils (TANs) and macrophages (TAMs). A pro-metastatic role of cPLA2α and its end-products (PGE2, LTB4) has been demonstrated in several cancer models, including lung cancer, TNBC, HCC, and colorectal cancer (CRC) [40,41,42,43]. For example, cPLA2α knock-down in high cPLA2α-expressing TNBC cells with increased metastatic capacities, delays TNBC tumorigenesis and strongly decreases the number of lung metastasis nodules [42]. Importantly, in TNBC patients, elevated cPLA2α levels are correlated with tumor aggressiveness, and poor overall- and free-disease survival [42]. In accordance with these findings, administration of PGE2 potentiates the invasive capacities of CRC cells and their ability to form liver and lung metastases in preclinical models [41]. In contrast, the blockade of PGE synthase (PTGES), downstream of COX (Fig. 1), inhibits metastatic lung tumor growth by decreasing abundance of immunosuppressive cells, such as myeloid-derived suppressor cells (MDSCs) and TAMs, and restoring the anti-tumoral immune populations (e.g., CD8+ T and natural killer (NK) cells) [43]. In regards with TME-derived eicosanoids, it has been shown that LTB4 and PGE2 exert a crucial role in establishing the pre-metastatic niche in breast and liver cancers [44]. For example, the CD11b+Ly6G+ neutrophils’ enrichment of lung metastatic TME, observed before mammary cancer cells have infiltrated the tissue, and the subsequent increase in neutrophil-derived LTB4-E4 favor the high metastatic potential of mammary tumor cells [44]. Hence, specific-inhibition of LT synthesis with Alox5 inhibitor, suppresses neutrophil pro-metastatic activity, and thereby decreases metastatic incidence [44]. Interestingly, ALOX5 inhibitor reduces also the number of lung metastatic foci arising from HCC cells, similarly to the selective depletion of alveolar macrophages which are the predominant cells that produce LTB4 in this model [45]. In addition to mediate a dialog between tumor cells and TME cells, PGE2 is also involved in the crosstalk between lung resident mesenchymal cells and metastasis-infiltrating neutrophils, which stimulates mammary metastatic dissemination [46]. Indeed, lung resident mesenchymal cells increase lipid uptake and TAG storage in metastasis-infiltrating neutrophils through PGE2-dependent and -independent mechanisms [46]. These neutrophils produce TAG-loaded extracellular vesicles, which through a macropinocytic uptake, increase cell proliferation and metastatic colonization capacities of tumor cells in metastatic mammary cancer models [46]. In addition, in the context of bone metastases caused by mammary cancer cell dissemination, increased ARA secretion by osteoclasts and decreased lysophosphatidylcholine (LPC) secretion, both synergistically support proliferation and pro-metastatic features of tumor cells [47]. Therefore, treatment combining an inhibitor of ARA-derived PG and LT with LPC supplementation greatly reduces the mammary metastatic incidence and spreading in bone environment. Altogethe